Psychological Safety

What it is

Edmondson published "Psychological Safety and Learning Behavior in Work Teams" in Administrative Science Quarterly in 1999, as sole author. It is a multimethod field study of fifty-one work teams in a single manufacturing company.

What it found is more specific than the popular version. Team psychological safety predicted learning behaviour. Team efficacy did not, once psychological safety was controlled for. Learning behaviour mediated the path from psychological safety to team performance, and both team structures, meaning context support and leader coaching, and shared beliefs shaped the outcome.

Her own practical account, given in a public talk, has three steps: frame the work as a learning problem rather than an execution problem, acknowledge your own fallibility, and model curiosity by asking a lot of questions. She has developed the construct further since, including in The Fearless Organization in 2018.

In effect

Google's multi-year internal study of team effectiveness expected star talent to be the driver and found it was not. Its summary, that who is on a team matters less than how the members interact, is the sentence that carried the idea into general management vocabulary, and of the five dynamics it identified, psychological safety was placed underneath the other four.

That study is not the evidence for the construct. It is internal corporate research. It was not preregistered, the measures are self-report, and no independent replication of it exists. The workforce is unrepresentative in every direction that bears on the question: highly educated, well paid, in knowledge-intensive roles, and self-selected into a culture that already prizes intellectual confidence. Cite Edmondson 1999 for the construct and for what predicts what, and cite the Google work as what it is, a large internal study that popularised a finding it did not establish.

The construct arrives from another direction in Hasard Lee's account of the fighter pilot debrief, in which rank is suspended and a junior flight lead can call out a base commander. He names no researcher and does not use the term, and he independently states the mechanism, its fragility and the leader's disproportionate responsibility: the debrief works only because everyone takes ownership, it is an unstable equilibrium that one blame-shifter can collapse, and the maintenance falls hardest on the people who could hide behind status. That is a practitioner's account, uncited and unfalsifiable as written, and it is tested against consequences measured in aircraft and in people rather than against a scale. Two weak but independent kinds of evidence pointing the same way is a stronger position than either alone.

Nir Eyal's use is the interesting extension: he makes psychological safety the root cause of workplace technology overuse, on the grounds that people who cannot say the always-on norms are unbearable cope instead, and the coping looks like distraction.

What it does not say

It does not mean niceness. The construct is widely acknowledged, including by Edmondson, to have been flattened in popular reception into being nice to each other, which is close to the opposite of what the 1999 paper argues.

It does not rest on Google's research. That work is internal, self-report and never independently replicated, and this publication does not treat it as the proof.

It does not come with the mediation claim often attached to it. The assertion that psychological safety mediates the link between good workplace relationships and organisational learning arrives in at least one popular account with no study, no sample and no coefficient, and should not be repeated as a finding.

It does not generalise cleanly across cultures. The research base is largely questionnaire-based and correlational, and the work is rooted almost entirely in Western and mostly American organisations, so cross-cultural generalisation is an open question.

---